Image-Derived Input Functions on an Ultra-High Performance Brain PET Scanner: Minimizing the Carotid Partial Volume Effect

Image-derived input functions (IDIF) are non-invasive alternatives to arterial blood sampling for PET kinetic modeling. However, partial volume effects (PVE) limit their applicability in brain PET. The NeuroEXPLORER (NX), a next-generation PET scanner with unprecedented sensitivity and spatial resolution, may allow for better IDIF extraction from carotid arteries. Six studies were performed on the NX (three tracers: ${ }^{18} \mathrm{~F}$-FDG, ${ }^{18} \mathrm{~F}$-SynVesT-1, ${ }^{11} \mathrm{C}$-PHNO). Arterial input functions (AIF) were acquired. Standard motion correction (MC) and MC based on a markerless motion-tracking system (UMT) were compared. Common carotid (CC) and internal carotid (IC) arteries were manually segmented on an early summed image. Region centerlines were generated, and dilated to 1-6 mm diameter to evaluate PVE. Percent IDIF-to-AIF area under the curve bias was calculated at different time intervals to estimate IDIF ability to recover whole-blood AIF. Using AIF and IDIF, kinetic modeling was applied to brain regions to estimate tracer uptake/binding ($K_{\mathrm{i}}$ for ${ }^{18} \mathrm{~F}$-FDG, $V_{\mathrm{T}}$ for ${ }^{18} \mathrm{~F}$-SynVesT-1 and ${ }^{11} \mathrm{C}$-PHNO) and inflow $\left(K_{1}\right)$. The percent bias of IDIF-to-AIF-based estimates was calculated. Using this simple segmentation approach, good recovery of AIF AUC was obtained for small centerline diameters (% bias < 5-10% for 1-2 mm). Kinetic parameter bias was very small for FDG $K_{\mathrm{i}}$ (% bias < 5%) and larger for ${ }^{18} \mathrm{~F}$-SynVesT-1 and ${ }^{11} \mathrm{C}$-PHNO $V_{\mathrm{T}}$. UMT MC provided lower bias. Overall, the usually challenging IC artery was equivalent or superior to CC. These results highlight how the NX helps minimize PVE in IDIF extraction.